# Recipe User Guide
## What is a Recipe?
A Recipe in askEdgi is a saved, reusable analytical workflow. It captures not just the code (SQL/Python), but the entire context of an analysis, including the datasets used ("ingredients"), the specific queries, and any AI enrichments applied. Recipes allow you to automate repetitive tasks and standardise reporting across teams.
It captures datasets, queries, AI enrichments, and visualisations in a structured, reusable format. Recipes eliminate the need to recreate multi-step analyses manually.
When you execute a recipe, askEdgi applies the same logic to the current data in your workspace. Results appear in a new chat with all outputs preserved.
### When to Use Recipes
Use recipes when:
* You perform the same analysis regularly (weekly reports, monthly summaries)
* Your team needs consistent KPI calculations across departments
* You want to share proven analytical workflows with colleagues
* You need to standardise data quality checks or metadata audits
Recipes reduce errors, save time, and ensure everyone uses the same analytical logic.
### Recipe Components
Every recipe contains the following elements:
* **Name**: A descriptive title for the recipe
* **Description**: An explanation of the recipe's purpose and outputs
* **Ingredients**: Datasets or metadata objects required for execution (optional)
* **Steps**: Queries, enrichments, and transformations captured from the original analysis
* **Analysis Type:** Indicates whether the recipe operates on data or metadata
When you save a chat as a recipe, askEdgi captures these components automatically. You can edit them before saving.
### Creating a Recipe
There are two ways to create a recipe: converting an existing conversation or building one from scratch.
#### Method 1: Convert to a Chat
This is the easiest way to save a successful analysis.
1. Open the **My Chats** panel on the left side of the interface.
2. Select the conversation chat you want to save.
3. Click the **Save as Recipe** button.
4. The system automatically extracts the logic and queries into steps.
5. Enter a **Name**, **Description**, and **Ingredients** for the recipe.
6. Click **Save as Recipe**.
#### Method 2: Creating from Scratch
1. Navigate to the **Recipes** tab in the main menu.
2. Click **Create Recipe**.
3. **Define Ingredients**: Select the data objects (tables or files) required for this recipe. Ingredients are optional but recommended for workflows that depend on specific data structures.
4. **Add Steps**: For each step of the analysis:
1. Select the **Type of Analysis** (SQL, Python, or AI Function).
2. Enter the specific **query**, **code**, or **prompt**.
5. Click **Save as Recipe.**
### Understanding Ingredients
Ingredients define which datasets a recipe requires. However, ingredients are optional.
#### When to include ingredients:
* The recipe is designed for specific tables (example: service request analysis requires the service\_requests table)
* You want to enforce that certain datasets must be present
#### When to omit ingredients:
* The recipe applies generic logic that works with any dataset
* You want maximum flexibility for reuse across different data environments
#### From My Recipes
1. Open the Recipes section.
2. Navigate to **My Recipes** tab.
3. Locate the recipe you want to run.
4. Click the **Execute** button.
5. The recipe runs in a new dedicated chat.
#### From the Chat Interface
1. Click the Execute Recipe icon next to the Send button.
2. Paste the recipe token (if using a Marketplace recipe).
3. Click **Execute.**
4. Results appear in the chat and are saved as a new chat.
#### From the Marketplace (Public Edition)
* **Access the Recipe from Marketplace**
* Navigate to the Marketplace section where all published recipes are displayed.
* Click on a recipe card to open its Recipe Details screen.
* On this screen, users can view the recipe’s description, verification status, associated tags, and pricing details.
* Once reviewed, click on the **Get Recipe** button to proceed. A pop-up window will appear displaying the Recipe Token.
* **Copy the Recipe Token**
* Copy the Recipe Token displayed in the pop-up. This token uniquely identifies the selected recipe for execution within askEdgi.
* **Open the Execute Recipe Panel**
* In the chat interface, click on the **Execute Recipe** icon located beside the **Send** button.
* Or click on the “**+Add Recipe from Markeplace**” button in **My Recipe** tab.
* The Add Recipe from Marketplace pop-up will appear.
* **Paste the Recipe Token**
* Paste the copied Recipe Token into the input field within the My Recipe pop-up.
* **Execute the Recipe**
* Click on the **Execute** button to run the recipe.
* askEdgi will process the recipe and display the results directly in the chat interface.
* **View Executed Recipe in Chats**
* Once executed, the recipe and its output are automatically saved and can be viewed in the Chats section.
* This allows users to revisit or reuse the executed recipe as part of their ongoing workspace history.
### AI Enrichment in Recipes
Recipes can include AI functions to transform data. Supported functions include:
* **Sentiment Analysis**: Classifies text as Positive, Neutral, or Negative.
* **Intent Analysis**: Categorises the purpose behind text (e.g., "Support Request," "Complaint").
* **Text Classification:** Sorts data into custom categories like "Fraud/Not Fraud".
* **Calculated Columns:** Creates new fields based on mathematical or logical formulas.
### Sharing Recipes Internally (SaaS Editions)
Enterprise users can share recipes with colleagues through the Organisation tab.
#### Listing a Recipe in an Organisation
1. Navigate to **Recipes**.
2. Click the **Organisation** tab.
3. Click "**List to Organisation**".
4. Select a recipe from **My Recipes.**
5. Click **Submit.**
The recipe becomes visible to other users in your organisation. Ownership remains with you, but authorised users can execute it according to role-based access controls.
**Using Organization Recipes**
1. Navigate to **Recipes.**
2. Click the **Organization** tab.
3. Browse recipes listed by your colleagues.
4. Click **Execute** on any recipe you want to run.
The recipe runs in your workspace using your available datasets and connectors. Results are stored in your chat history.
### Publishing to the Marketplace (Public Edition Only)
Public edition users can publish recipes to the Marketplace for community use and monetisation.
**Step 1: Prepare the Recipe**
Ensure your recipe is complete, well-documented, and tested. Review the name, description, and steps for clarity.
**Step 2: List to the Marketplace**
1. Navigate to **My Recipes**
2. Select the recipe
3. Click "**List to Marketplace**"
4. Click the "**Marketplace Edit**" link
**Step 3: Complete the Marketplace Edit Form**
**Basic Information:**
* Recipe Name: Clear, descriptive title
* Recipe Slug: Unique identifier for the recipe URL (auto-generated, editable)
* Description: Detailed explanation of purpose, outputs, and use cases
* Recipe Guide: Step-by-step instructions for execution
**Seller Information:**
* Seller Support: Contact email or support link
* LinkedIn Profile: Optional professional profile link
**Pricing:**
* Pricing Type: Free or Paid
* Per Execution Price (USD): If Paid, specify price per execution
**Categories:**
* Industry: Select relevant sector (Retail, Finance, Healthcare, etc.)
* Application: Identify business application (Product Returns, Customer Sentiment, etc.)
* Others: Additional applicable categories
**Tags:**
* Enter keywords separated by commas to improve discoverability
**Example**: returns, sentiment, quality
**Step 4: Request to Publish**
* Click "**Request to Publish**". The recipe is submitted for review by the OvalEdge admin team. Admins verify quality, completeness, and compliance before approval.
**Step 5: Approval and Publication**
* Once approved, the recipe becomes publicly available in the Marketplace. Other users can subscribe, execute, and provide feedback. If you set a paid pricing model, earnings accrue to your account based on execution counts.
#### **Managing Published Recipes**
**Viewing Earnings**
1. Click the **profile** icon
2. Select "**Manage Subscription**"
3. Navigate to the Earnings tab
The Earnings tab displays revenue generated from recipe executions. Data is structured by recipe type, quantity, price, and total earnings. A historical chart shows performance across months.
**Revoking a Recipe**
If you need to withdraw a published recipe:
1. Access the Marketplace dashboard (creator view)
2. Locate the recipe
3. Select the "**Revoke**" action
The recipe is removed from public discovery. New consumers cannot subscribe or execute it. Existing executions already purchased may continue to function.
### Recipe Execution by Edition
Recipe behaviour varies by deployment edition. Understand which recipes you can create and execute based on your environment.
#### Public Edition
**Capabilities:**
* Create recipes in the workspace
* Execute recipes on uploaded or public datasets
* Full access to the Public Marketplace
* Publish and monetise recipes
**Constraints:**
* No enterprise connector access
* Limited to public data and uploaded files
#### SaaS – Data Analytics
**Capabilities:**
* Full internal recipe creation
* Execute all recipe types (data and metadata)
* Share recipes via the Organisation tab
* Optional access to Public Marketplace recipes (policy-controlled)
**Scope:**
All connectors, uploaded files, and catalog tables are supported
#### SaaS – Metadata Analytics
**Capabilities:**
* Create metadata-focused recipes
* Execute recipes operating through the OvalEdge (-1 connector)
* Share recipes internally
* Optional access to Public Marketplace recipes (policy-controlled)
**Constraints:**
* Only metadata recipes are supported
* No data-level enrichments
#### On-Prem Edition
**Capabilities:**
* Create internal metadata analytics recipes
* Execute recipes using the OvalEdge (-1 connector)
* Share recipes internally (optional, policy-controlled)
**Constraints:**
* No external AI enrichments
* No public recipe sharing
* No Marketplace access
* Metadata recipes only
### Special Case: Recipes Without Ingredients
Some recipes operate without requiring specific datasets. These are typically metadata-focused workflows.
**Example**: Data quality checks across all tables in the catalog
In this case:
* No ingredients are defined in the recipe
* Execution runs against all tables in the workspace or catalog
* Results depend on which objects are present during execution
This pattern is common for governance checks, profiling summaries, and catalog audits.
### Recipe execution: Common failure reasons
* Required datasets are missing from the workspace
* Insufficient permissions to access catalog objects
* Column names or structures differ from the recipe's expectations
* Workspace compute limits exceeded
## Best Practices
**Naming Conventions**
Use clear, descriptive names that explain what the recipe does. Avoid generic names like "Analysis 1" or "Report."
**Good examples:**
* "Monthly Sales Summary by Region"
* "Customer Sentiment Analysis with Risk Classification"
* "Data Quality Check for Orders Table"
**Documentation:**
Write descriptions that help others understand the recipe's purpose, required inputs, and expected outputs. Include guidance on when to use the recipe and what insights it provides.
**Testing Before Sharing**
Execute the recipe on test data before sharing it with your organization or publishing to the Marketplace. Verify that all steps produce correct results and that visualizations are accurate.
**Version Control**
If you modify a recipe significantly, save it as a new recipe with a version indicator in the name (example: "Customer Analysis v2"). This preserves the original workflow while allowing experimentation.
**Metadata Recipes**
When creating metadata recipes, document which catalog objects or metadata domains the recipe targets. This helps users understand whether the recipe applies to their governance needs.
### Troubleshooting Recipe Execution
**The recipe fails immediately**
* Cause: Required datasets are not in the workspace
* Solution: Add the necessary catalog objects or files to your workspace before executing the recipe
**The recipe produces unexpected results**
* Cause: Column names or structures differ from the original analysis
* Solution: Review the recipe steps and verify that the datasets in your workspace match the expected schema
**Recipe times out or exceeds limits**
* Cause: Large datasets or complex joins exceed workspace capacity
* Solution: Upgrade the workspace container or simplify the recipe by breaking it into smaller steps
**The recipe cannot be executed in my edition**
* Cause: The recipe requires connectors or features not available in your edition.
* Solution: Verify that the recipe is compatible with your deployment model (Public, SaaS Data Analytics, SaaS Metadata Analytics, or On-Prem)
## Example of Recipe
### Creating a Recipe with a Real Use Case
#### Example Scenario
**Business Question:**
“How are customers feeling about our support tickets, and which issues need immediate attention?”
This example walks through creating a Customer Support Sentiment Analysis recipe using a real dataset and AI enrichment.
#### Step 1: Start with an Analysis (Chat-Based)
Assume you already have a table in your workspace:
Table: support\_tickets\
Columns:
* ticket\_id
* created\_date
* customer\_comment
* priority
* category
You run the following steps in a chat.
#### Step 2: Basic Data Selection
```
SELECT
ticket_id,
created_date,
customer_comment,
priority,
category
FROM support_tickets
WHERE created_date >= current_date - INTERVAL '30 days';
```
This filters tickets from the last 30 days.
#### Step 3: AI Enrichment – Sentiment Analysis
Apply Sentiment Analysis on the customer\_comment column.
**AI Function**: Sentiment Analysis
**Input Column**: customer\_comment
**Output Column**: sentiment
**Values**: Positive / Neutral / Negative
#### Step 4: AI Enrichment – Intent Classification
Add an **Intent Analysis** step to understand why customers raised tickets.
**AI Function**: Intent Analysis
**Input Column**: customer\_comment
**Output Column**: intent
**Example Values:**
* Complaint
* Support Request
* Feature Request
* Billing Issue
#### Step 5: Derived Insight (Optional)
Create a calculated column to flag **high-risk tickets.**
```
CASE
WHEN sentiment = 'Negative' AND priority = 'High' THEN 'High Risk'
ELSE 'Normal'
END AS risk_flag
```
#### Step 6: Save the Chat as a Recipe
Click **Save as Recipe** and fill in:
* **Recipe Name**:\
Customer Support Sentiment & Risk Analysis
* **Description**:\
Analyzes customer support tickets from the last 30 days, classifies sentiment and intent using AI, and highlights high-risk tickets requiring immediate attention.
#### Step 7: Define Ingredients
**Ingredients (Included):**
* support\_tickets (table)
**Why include ingredients?**
This recipe depends on a specific table structure and should only run when that dataset is available.
#### Step 8: Review Recipe Components
When saved, the recipe contains: